Dirt Diggers Digest – Issue 14, February 2023

This post will address a frequent question posed at crop schools and field days in formal Q&A’s as well as side-bar discussions, and that is, “Why are results from different soil labs often so variable on the same or similar samples I submit?”

It is a very reasonable question, and one that the whole soil testing industry spends thousands of hours and dollars on each year to address. The answer to the question, at least gaining an understanding of the basis of the answer, lies in two critically important factors:

1. Controlling sampling errors that carry through lab analyses by:
   1. Making sure field samples are representative of the area you are managing.
   2. Helping minimize laboratory subsampling errors with proper sample prep.
      
      
2. Choosing appropriate labs and methods and understanding inherent analytical errors by:
   1. Choosing a regional lab that routinely deals with samples similar in character to those you are submitting, and requesting a regionally appropriate test method.
   2. Realizing that specific soil test methods have inherent limitations on precision, and variability within that typical error is entirely normal.

I have had to pay particular attention to the details around this question for a large part of my career, especially from 2009 to 2019 as I served as the coordinator of the North American Proficiency Testing Program (NAPT Program) for the Soil Science Society of America. The NAPT Program is a soil testing laboratory third-party proficiency evaluation program and an important part of the quality control and assurance efforts of over 150 soil testing labs across the US, Canada, Mexico and several international locations. Gleaning from my experience working with the very best labs in North America, let me share some things that I have learned that address the above mentioned factors.

Sampling Error

Let’s begin with controlling variability in lab results arising from in-field sampling. Ensuring a good result begins on the ground in the field. Proper soil sampling takes only a little bit of focused effort, but will return much better and more consistent results from which to make sound management decisions. Cutting the corners on soil sampling such as jumping out of the truck on one edge of the field, using your knife to loosen some surface soil and bagging that up for submission to the lab (not an uncommon scenario) may or may not (more likely) be representative of general field conditions, and will never provide a reproducible baseline result to compare to going forward.

The best method for soil sampling is to create a “physical average” of soil conditions by compositing multiple, uniform volume soil samples from randomly determined locations within the field. There are several ways to accomplish this, but using a core-type soil sampler, and using a crisscross sample pattern across the field is perhaps the simplest, yet highly effective method. Care should be taken to sample and test dissimilar areas (i.e., areas having different slope, soil type, different previous or current crop from other areas of the field, etc.) separately.

Keep in mind that fertility and management recommendations are based on the top 12 inches of soil, so sample over that entire depth, and thoroughly mix samples from 10 to 20 locations in similar areas of the field together. This will form a “physical average” sample over the area, and should allow for a reproducible result for future comparison. One can further help the lab reduce error by breaking down the sample as much as possible and mixing as completely as possible before sending the sample to the lab. This ensures that the lab won’t inadvertently subsample from a portion that is not adequately mixed with the rest of the sample. (NOTE: a well-mixed composite sample can represent up to 25 acres, so more individual composite samples are needed for larger areas).

The same principle applies if one desires to send the same sample to separate labs for comparison. A matched subsample must be split from the same composite sample to ensure the two labs have as uniform a sample as possible from which to work. Otherwise, sending two grab samples from the field without thorough compositing or carefully splitting a mixed composite sample, will introduce error that will confuse the results.

Sample preparation before the lab ever sees the sample, is the most critical step in producing good results. Shoddy sample preparation can result in more than 100% variability for some soil properties, and hence, a lab result that will not likely reflect the needs of the field, or ever be a reproducible baseline result to which future sample analyses can be compared

Selecting the Appropriate Soil Test Method

When one wishes to compare the results of one lab to another, or for some reason, sends samples from the same field to a different lab than was used in the past, one must ensure that the same analysis methods are requested. Assuming good sampling and preparation as discussed previously, then comparable results from two different labs depends on requesting the same analytical method for a particular nutrient or soil property from each lab. For instance, there are more than seven established methods for determining plant available phosphorus, namely, the Olsen Method, Bray 1:7, Bray 1:10, Mehlich 1, Mehlich 3, Morgan, Kewlona and other minor methods. Laboratories will default to the analytical method most often used in the area the lab operates, so be sure to specify the same method when comparing lab results.

There are no simple conversions between the results from these various tests as each uses a different chemical extractant which has been separately calibrated to plant response and carries its own specific interpretation and associated management recommendation. Hence, apples-to-apples comparisons of one lab result to another must be based on the same analytical method. In a future post, I will go through which methods are most appropriate for Utah and the Intermountain West.

Inherent Analytical Error

Each analytical method has inherent variability associated with it. This is due to a number of factors such as the precision of the machine or instrument used to measure a given soil property, the difficulty in extracting a given nutrient or element from a soil, the stability of chemical reactants employed, the skill of the lab staff, and many other factors. Most soil testing labs in the U.S. subscribe to one of two proficiency testing programs that help them track their analytical accuracy and precision and adjust their quality controls as needed.

Despite the quality control efforts of labs, there are still some analytical methods that are inherently variable. Having worked with hundreds of labs over the past decade, I have had opportunity to review and compare analytical error over much of the soil testing industry. The following graphic is from a recent journal article that I co-authored about this subject.

The bars in Figure 1 indicate the precision of a number of soil test methods over a large number of laboratories conducting that same method on controlled standard soils. The numbers above the bars indicate the number of labs submitting data for a given analytical method. The precision score is an average of three years of collected quarterly data (12 sample periods) over that group of labs. Over half of the methods can be controlled within 10% variation in result, which is the desirable limit of analytical variation. Many methods, however, commonly exceed 10% variation, with some as high as 15-20%.

It is nice to know that for our most important nutrients (N, P and K) and key soil properties (pH, water retention, texture, etc.) the variation is typically less than 10% across the industry. However, even at that level of analytical error control, it would not be uncommon to find results for plant available P for instance (about 8% typical variability) to span a range from sufficient to marginally deficient for two separate subsamples of the same soil. Couple that with poor field sampling or poor sample preparation, and the variability in lab results could be quite large between samples from the same field.

Final Note

Even though there is a measure of unavoidable analytical error (more for some analyses than others) one must understand that almost all public and private soil testing labs spend hundreds of dollars each year to regularly monitor and adjust internal procedures, and subscribe to third-party proficiency programs to evaluate and validate their results. In almost all cases, the variability seen in processing samples from the same field over time, or the difference in results between two labs on the same/similar soil sample, is far more likely a case of field sampling error than laboratory analytical error, or unknowingly trying to compare results of two different analytical methods for the same soil property.

Author

Grant Cardon